ExperimentalFor this study, a single-grain slab of a commercial nickel base alloy was received in the as cast condition. The composition of the main alloying elements is listed in Table 1. From this slab, two bars with a diameter of about 20 mm were machined by spark erosion parallel to the direction of solidification and several pieces 15 mm in length were cut for the following heat treatments. l HT1: as cast, i.e., no heat treatment. l HT2: as cast + aged 6 h at 1080 C, i.e., 0 % solutioned and final aged. l HT3: solutioned 0,5 h at 1213 C + aged 6 h at 1080 C, i.e., partial solutioned and final aged. l HT4: solutioned 0,5 h at 1260 C + aged 6 h at 1080 C, i.e., 100 % solutioned and final aged. For microstructural investigations, the specimens were mounted, ground and polished, and finally etched by the so-called molybdenum acid. This type of reagent which attacks the c¢-phase is called negative etching. Quantitative image analysis was carried out in order to determine the c¢-volume fraction and the c¢-average size.Chemical microanalyses were carried out on a scanning electron microscope LEO 435 VP equiped with an energy dispersive spectrometer (EDS) PGT IMIX-PC. All observations and analyses were performed under 15 kV. Some X-ray maps were recorded to look for the composition of precipitates as described later. However, most of the analyses were carried out to characterize the distribution of the alloying elements at the scale of the dendritic structure. For this purpose, composition was estimated in spot mode at points located along a grid, 50 lm apart each other. Measurements have been made on 300 points for sample HT1 and 120 points in the case of sample HT4. The counting time in each point was set to 60 s for both grids. The EDS data in each point was then treated for ZAF correction using standards of pure elements, except for carbon which was not analyzed. The total species content estimated in each point, as well as the so-called ªgoodness of fitº (GOF) which characterizes the quality of the analysis, are reported on Figure 6 for each of the grids. It is first seen that the total content tends to decrease with the rank of the point analyzed. This trend is related to a continuous shift of the beam intensity during the analysis, i.e., with time. For this reason, use has been made afterwards of solute content after normalization (i.e., correction of the sum to 100 %). Concerning the quality of the analyses, it is seen that in most cases the GOF had a very low value. Counting with value higher than unity have been excluded for further analysis. Fig. 6. Correlation between silicon and manganese in the same area as Figure 3 below the peritectic temperature.